Home-based rehabilitation can serve as an adjunct to in-clinic rehabilitation, encouraging users to engage in more practice. However, conventional home-based rehabilitation programs suffer from low adherence and high drop-out rates. Wearable movement sensors coupled with computer games can be more engaging, but have highly variable adherence rates. Here we examined characteristics of user adherence by analyzing unsupervised, wearable grip sensor-based home-hand rehabilitation data from 1,587 users. We defined three different classes of users based on activity level: low users (<2 days), medium users (2 - 9 days), and power users (> 9 days). The probability of using the device more than two days was positively correlated with first day game success (p = 0.91, p<. 001), and number of sessions played on the first day (p = 0.87, p<. 001) but negatively correlated with parameter exploration (total number of game adjustments / total number of sessions played) on the first day (p = - 0.31, p= 0.05). Compared to low users, power users on the first day had more game success (65.18 ± 25.76 %vs. 54.94 ± 30.31 %,p <. 001), parameter exploration (25.47 ± 22.78 % vs. 12.05 ± 20.56 %, p <. 001), and game sessions played (7.60 ± 6.59 sessions vs. 4.04 ± 3.56 sessions, p <. 001). These observations support the premise that initial game success which is modulated by strategically adjusting parameters when necessary is a key determinant of adherence to rehabilitation technology.